Unit 4, Flex
Topic | Description | Link |
---|---|---|
Lesson | Part I. Working With Time Series Data | Here |
Part II. Rolling Statistics | Here | |
Part III. Autocorrelation | Here | |
Part IV. Decomposition | Here | |
Practice | Part V. Time Series Practice | Here |
Lesson | (Bonus) Part VI. Time Series Forecasting Models | Here |
Solutions | Sample solutions for lesson prompts and practice activity | Here |
Data | 10 datasets used in lesson samples and practice activities | Here |
After this lessons, students will be able to:
- Explain how time series data are distinct and how to account for that.
- Create rolling means and plot time series data.
- Perform autocorrelation on sample time series data.
- Decompose time series data into trend and residual components.
- Bonus: Explain the benefits and drawbacks of ARIMA as compared to other methods.
Before this lesson(s), students should already be able to:
- Manipulate data with Pandas.
- Create basic data visualizations with Matplotlib and Seaborn.
- Define and demonstrate techniques for correlating data points.
TOTAL (190 min) I. Foundations (190 min)
- Part 1: Time Series Data (20 min)
- What is Time Series Data?
- Examples of Time Series Data and Domains That Find it Useful
- Trends and Seasonality
- Preprocessing Time Series Data With Pandas
- Processing Date Information as datetime Objects
- Filtering by Date With Pandas
- Differencing/Lagging
- Practice (15 min)
- Part 2: Rolling Means and Plotting Time Series Data (20 min)
- What is a Trend?
- Visualizing Trends
- Computing Aggregates With Pandas' .resample() Function
- Computing Rolling Averages With Pandas
- Other Pandas Window Functions (Rolling Sum, Rolling Max, Diff)
- Exponentially Weighted Windows
- Plotting Time Series Data
- Practice (15 min)
- Part 3: Autocorrelation on Time Series Data (25 min)
- Define Autocorrelation
- Examples of Autocorrelation
- Computing Autocorrelation
- Using the Pandas autocorr() Function Calculating and Plotting ACF and PACF Using StatsModels
- Practice (15 min)
- Part 4: Decomposing Time Series (15 min)
- Decomposing Time Series Into Trend, Seasonality, Cyclical, and Residual Components
- Using the .seasonal_decompose() Function
- Plotting the Seasonal, Trend, and Residual Components of a Time Series
- Practice (15 min)
- Closing and Recap (5 min) II. Bonus (30 min)
- Time Series Forecasting Models (25 min)
- What are Time Series Models?
- Properties of Time Series Models
- Assumption of Stationarity
- Compare/Contrast Time Series Forecasting Models
- Closing and Recap (5 min)
If you are interested in more resources, check out the following:
- In Pandas' datetime library, search for more information on .dt here.
- For additional review of these concepts, see some inspiration from the Python Data Science Handbook.
- There are lots of additional tutorials on ARIMA models out there; here is a good one.